aggregate_metrics.json

{
  "test_macro_f1_mean": 0.6937477912021662,
  "test_macro_f1_std": 0.004536414285693068,
  "test_macro_f1_text": "0.6937 ± 0.0045",
  "test_accuracy_mean": 0.6952117863720074,
  "test_accuracy_std": 0.003987225615950428,
  "test_accuracy_text": "0.6952 ± 0.0040",
  "test_macro_precision_mean": 0.6955007583676345,
  "test_macro_precision_std": 0.0053962198757247,
  "test_macro_precision_text": "0.6955 ± 0.0054",
  "test_macro_recall_mean": 0.6935251566187915,
  "test_macro_recall_std": 0.004259167819556978,
  "test_macro_recall_text": "0.6935 ± 0.0043",
  "dev_macro_f1_mean": 0.7248417464601334,
  "dev_macro_f1_std": 0.00291168417848945,
  "dev_macro_f1_text": "0.7248 ± 0.0029",
  "task": "ViFactCheck-full-context",
  "dataset": "ViFactCheck",
  "model_key": "cafebert",
  "model_name": "CafeBERT",
  "base_model": "uitnlp/CafeBERT",
  "seeds": [
    22,
    42,
    202
  ],
  "representative_seed": 202,
  "selection_rule": "maximum development Macro-F1; seed ascending tie-break",
  "split_policy": "merged_stratified_80_10_10",
  "split_seed": 42,
  "max_length": 256,
  "epochs": 3,
  "effective_batch_size": 8
}